Work Package Prompt: WP02 – Analysis-currency gate generalized to any analyze-declaring family

Problem

The stale-analysis → re-analyze currency gate is hard-keyed to the literal software-dev family at three runtime sites, so an analyze-declaring custom type never gets it:

src/runtime/next/runtime_bridge_io.py:1543 — if mission_family == MISSION_TYPE_SOFTWARE_DEV and step_id == ANALYZE_STEP_ID: (computes ANALYSIS_CURRENCY_FACT).
src/runtime/next/runtime_bridge.py:652 — if board.action == _mapping._STEP_IMPLEMENT and ctx.mission_type == MISSION_TYPE_SOFTWARE_DEV: (reissues analyze on the finalized-board override; decide path).
src/runtime/next/runtime_bridge_query.py:144 — same, preview/query path.

Goal

Generalize all three to key on the mission's resolved runtime DAG declaring an analyze step.

Constraints (binding)

action_sequence is specify, plan, tasks, implement, review — it has NO analyze — yet software-dev HAS an analyze DAG step. Keying on action_sequence would silently drop the gate for the one family that has it today (an NFR-001 regression). The injected probe analysis_currency_for (src/specify_cli/cli/commands/next_cmd.py:97-141) is already family-agnostic and injected unconditionally — do not change the injection wiring.

declares an analyze step among built-ins. Keep tests/runtime/test_runtime_seam.py and test_next_board_authority.py green with 0 diffs.

Key on the DAG step SET, never action_sequence membership. Built-in software-dev's
Byte-identical for software-dev / research / documentation / plan. Verified: only software-dev

Subtasks

ANALYZE_STEP_ID in {s.id for s in <frozen template>.steps}, loading the frozen template via the existing loader (runtime_bridge_engine._load_frozen_template / load_mission_template_file over run_dir / "mission_template_frozen.yaml"). ANALYZE_STEP_ID is runtime_bridge_cores.py:106.

(the step is only ever issued when the resolved runtime declares it, so this already encodes the key).

with mission_declares_analyze_step(ctx.run_dir) (ctx.run_dir is on DecideNextContext; the override is already gated on _run_is_untouched(ctx.run_dir) so the template is frozen).

into _build_finalized_override_query_decision from its caller _query_dispatch_decision.

tests/runtime/_next_mission_scaffold.py; red-first tests in tests/runtime/test_analyze_step.py: (a) currency fact populated at analyze for the custom family + software-dev unchanged + a no-analyze family (research) absent (FR-003/FR-005); (b) decide-path reissue of analyze on a stale analysis-report.md + a fresh-analysis control that proceeds to implement (FR-004/FR-005); (c) preview-path preview_step == "analyze" on a stale report + no-analyze control (FR-004).

T004 — Add mission_declares_analyze_step(run_dir) -> bool in runtime_bridge_cores.py:
T005 — runtime_bridge_io.py:1543: drop the family clause → if step_id == ANALYZE_STEP_ID:
T006 — runtime_bridge.py:652 (decide path): replace ctx.mission_type == MISSION_TYPE_SOFTWARE_DEV
T007 — runtime_bridge_query.py:144 (preview path): same predicate; thread run_ref.run_dir
T008 [P] — Add an analyze-declaring non-software-dev custom scaffold in

Test economy (DIRECTIVE_041 / development-assist-test-cleanup)

One focused test per site, each paired with its no-op/no-analyze control for non-vacuity. Do not duplicate test_runtime_seam.py's golden-parity assertions; strengthen the existing analyze-step tests in place where the new family fits rather than cloning whole test bodies. A red-first demonstration that is not worth keeping goes in the implementer report, not the commit. Provide a table: test → fold → regression caught → existing coverage.

Definition of done

report, decide and preview alike; software-dev/research/documentation/plan byte-identical.

An analyze-declaring custom family refuses implement and re-issues analyze on a stale/missing
ruff/mypy clean, complexity ≤ 15; NFR-001 golden-parity suites green.

Spec Kitty

Kanban Board